Epitaxial growth of full range of compositions of (1 1 1) PbZr1- xTixO3 on GaN

Contents lists available atScienceDirect

Journal of Crystal Growth

Epitaxial growth of full range of compositions of (1 1 1) PbZr

1-x

Ti

x

O

3

on

GaN

Lin Li

, Zhaoliang Liao

, Minh Duc Nguyen

, Raymond J.E. Hueting

, Dirk J. Gravesteijn

,

Evert P. Houwman

, Guus Rijnders

, Gertjan Koster

a_{MESA+ Institute for Nanotechnology, University of Twente, Enschede, the Netherlands}

b_{National Synchrotron Radiation Laboratory, University of Science and Technology of China, Hefei 230026, Anhui, China}

A R T I C L E I N F O

Communicated by A. Ohtomo Keywords:

Pulsed Laser deposition Oxides

Perovskites Nitrides

Physical vapor deposition processes Ferroelectric materials

A B S T R A C T

Integrating functional complex oxides with conventional (“non-oxide”) semiconductors emerges to be an im-portant researchfield and has been attracting great interest. Because of their superior intrinsic material prop-erties, such as a relatively high dielectric constant and polarization, the utilization of PbZr1-xTixO3(PZT)

ma-terials as a dielectric layer is expected to greatly improve the performance of the GaN high electron mobility transistor. The functional PbZr1-xTixO3exhibits quite different crystal structures and consequently physical

properties depending on the composition. In this work we report the growth of full range of compositions of PZT films on MgO buffered GaN substrates. Besides revealing the temperature effect on phase formation and surface morphology, we demonstrated the strong effect of composition on the growth: pure (1 1 1) phase is formed in Ti-rich PZT (x > 0.48) while pyrochlore impurity phase is found in Zr-Ti-rich PZT (x < 0.48). By introducing an ultrathin Ti-rich PZT seed layer, we are able to achieve epitaxial growth of Zr-rich PZT. The epitaxial PZTfilms of different composition all exhibit good ferroelectric properties, showing great promise for future GaN device applications.

1. Introduction

The semiconductor GaN is an important material with many po-tential applications such asfield effect transistor (FET) in high power and high frequency devices due to its direct, wide band gap of 3.45 eV at room temperature and high chemical and mechanical stability[1,2]. Furthermore, GaN-on-silicon offers many advantages such as good electrical and thermal conductivity, large size availability, mass pro-duction, and significantly low cost for GaN optoelectronic and micro-electronic devices. The utilization of ferroelectric PbZr1-xTixO3(PZT)

materials as a dielectric layer is expected to further improve the per-formance of GaN/AlGaN high electron mobility transistor (HEMT). The superior material properties including high permittivity, excellent fer-roelectric, piezoelectric, and electromechanical properties[3]are ex-pected to make PZT an ideal material for improving the tradeoff be-tween the breakdown voltage and the specific on-resistance of the GaN/ AlGaN HEMT as well as realizing resonators, non-volatile FETs in a GaN platform[4–7].

A lot of effort has been spent to achieve high quality PZT films on GaN[8–14]. However, the highly incompatible crystal lattice, different lattice constant, and chemical reactivity between PZT and GaN hinder

the epitaxial growth of PZT[8–13]. The employment of buffer layers has been found to improve the film quality. For example, with an atomicallyflat ultrathin MgO buffer layer, we previously realized the epitaxial growth of highly crystalline PZT (x = 0.48)films with ex-cellent ferroelectric properties, which opens the possibility to practi-cally use advanced oxide materials in GaN HEMTs[13]. Depending on the composition the PZT possesses quite a different crystal structure and consequently physical properties [3,15]. A morphotropic phase boundary (MPB) at × = ~ 0.48 separates the rhombohedral (x < 0.48) from tetragonal phase (x > 0.48). The direction of fer-roelectric domains also depends on the composition. At the MPB, PZT exhibits the highest dielectric constant. The polarization direction is along the pseudocubic [0 0 1] direction in tetragonal PZT (x > 0.48) while it is along pseudocubic (1 1 1) direction for rhombohedral PZT (x < 0.48). Different compositions are suitable for different types of applications, hence it is highly desired to study the epitaxial growth of different compositions of PZT on GaN. In contrast to the growth of PZT on typical perovskite substrates such as SrTiO3 and GdScO3 where

growth is nearly composition independent [16,17], the growth of a different composition of PZT on MgO/GaN is non-trivial. Our results demonstrate that epitaxial growth is favored when × > 0.48 while

https://doi.org/10.1016/j.jcrysgro.2020.125620

Received 15 October 2019; Received in revised form 8 March 2020; Accepted 19 March 2020 E-mail address:g.koster@utwente.nl(G. Koster).

Journal of Crystal Growth 538 (2020) 125620

Available online 21 March 2020

0022-0248/ © 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

the non-ferroelectric pyrochlore phase will form when × < 0.48. By introducing PZT (x = 0.8) as an additional seed layer, we further realized the epitaxial growth of Zr-rich PZT (x < 0.48)films. In other words, we demonstrate that the complete PZT family can be epitaxially grown on MgO buffered GaN substrates. The physical properties of these (1 1 1) PZT layers on GaN are found to strongly depend on the composition, paving a path to diverse device applications.

2. Experimental

The PZTfilms were grown by pulsed laser deposition (PLD) on MgO buffered GaN using a ceramic PZT target. In order to compensate the loss of volatile Pb during the PLD deposition, an excess 10% Pb was added in the target [18,19]. The GaN was epitaxially grown on Si (1 1 1) substrates by metal organic chemical vapour deposition (MOCVD) with multiple Al1-xGaxN buffer layers to release the strain

from Si. Detail on GaN/Si substrates can be found in Supplemental in Ref.[13]. The MgOfilms were grown on GaN substrates by pulsed laser

deposition (PLD) from a stoichiometric MgO ceramic target at afluence of 5.6 J cm−2and repetition rate of 5 Hz. The growth temperature was maintained at 650 °C. To avoid oxidation of GaN, the MgO wasfirst grown in high vacuum at 10-7mbar for 2 min (500 pulses), then grown further at a 10-3_{mbar oxygen partial pressure. Reflection high energy}

electron diffraction (RHEED) was used to monitor in-situ the growth of MgO and to accurately control its thickness. More details of the growth of MgO can be found in Ref. 13. Regarding the PZT growth, thefluence and oxygen partial pressure were 2 J/cm2_{and 0.1 mbar, respectively.}

The growth temperature ranged from 540 °C to 680 °C. The x-ray dif-fraction (XRD) was performed using PANalytical-X 'Pert materials re-search diffractometer (MRD) in high resolution mode. The morphology was characterized by atomic force microscopy (AFM). Given that the 2 dimensional gas which is buried underneath top insulating GaN/ Al0.2Ga0.8N layers can be switched off by up polarization of PZT layer

and thus full P-E loop cannot be measured, 2 dimensional gas could not be used as bottom electrode. Instead, an in-plane capacitor configura-tion was used for ferroelectric measurement. The in-plane electrodes for ferroelectric measurement were fabricated with standard photo litho-graphy, Ti and Pt sputter deposition and lift-off. The ferroelectric properties were measured using the AixAcct TF-2000 Analyzer at a frequency of 1 kHz.

3. Results

First, the effect of growth temperature on the quality of PZT (x = 0.48) was investigated.Fig. 1a shows the XRD scan of 1.5 µm PZT (x = 0.48) on 115 nm MgO buffered GaN. At a relatively high growth temperature of 680 °C, only non-epitaxial (1 0 1)/(1 1 0) phase was observed. The preferred formation of {1 1 0} phase at high temperature is presumably due to a difference in nucleation energy of (1 1 1) PZT compared to {1 1 0} PZT, apparently causing a higher temperature to favor the formation of the {1 1 0} phase[20]. When the growth tem-perature was decreased to 630 °C, the {1 1 0} impurity phase was suppressed and a large amount of epitaxial (1 1 1) phase was formed. The lower growth temperature also resulted in less crystallinity of (1 0 1) and (1 1 0) phase, broadening both (1 0 1) and (1 1 0) peaks.

Fig. 1. XRDθ-2θ scan of PZT (x = 0.48) films on MgO buffered GaN/Si substrates grown at different temperatures. Data of pure GaN/Si substrate (black) is shown for comparison. (b) The rocking curves of the PZT (1 1 1) peaks forfilms grown at 570 °C (black and red curves) and 540 °C (green curve). (c) The surface morphology by AFM of 1.5 µm thick PZTfilms grown at different temperatures. The scale bar is 2 µm. The thickness of MgO buffer layer for 1.5 µm and 0.75 µm thick PZTfilms are 115 nm and 11 nm respectively.

Fig. 2. XRDθ-2θ scans of 750 nm PZT films with different compositions on MgO buffered GaN substrates. The thickness of MgO buffer layer is 0.5 nm.

Therefore, (1 0 1) and (1 1 0) merge to behave like single broad peak. At an even lower temperature of 570 °C, only the (1 1 1) orientation was observed. Similar to a previous report[13], the epitaxial growth of PZT was not affected by the thickness of MgO buffer layer. If reducing

the MgO buffer layer thickness to 11 nm, high quality of epitaxial PZT was still obtained (see red curve inFig. 1a). Surprisingly, single (1 1 1) orientation was still observed even at the relatively low growth tem-perature of 540 °C.Fig. 1b shows the rocking curve of PZT (1 1 1) peak

Fig. 3. (a) XRDθ-2θ scans and (b) Phi-scans of (4 2 0) peaks of 750 nm PZT(x = 0.2), PZT(x = 0.48) films on PZT(x = 0.8)/MgO/GaN (with a 5 nm PZT (x = 0.8) seed layer) and 750 nm PZT(x = 0.8)film on MgO/GaN. In (b) GaN (1–24) is shown for comparison. (c) AFM surface morphology of the PZT films.

Fig. 4. (a) Measured P-E loops, the inset in the PZT (x = 0.2) panel shows the in-plane electrode configuration; (b) extracted remnant polariza-tion (Pr); (c) extracted coercivefield (EC) of PZT

(750 nm)films of different composition on MgO (0.5 nm)/GaN substrates. For PZT (x = 0.2) and PZT (x = 0.48)films growth, an additional 5 nm PZT (x = 0.8) seed layers were grownfirst on MgO/GaN.

of different films grown at 570 °C and 540 °C. The full width at half maximum (FWHM) of 1.5 µm PZT on MgO(115 nm)/GaN, 0.75 µm PZT on MgO(11 nm)/GaN both grown at 570 °C, and 0.75 µm PZT on MgO (11 nm)/GaN grown at 540 °C is 0.59°, 0.62° and 0.61°, respectively. Therefore, lowering the growth temperature to 540 °C didn’t seem to affect the film quality, making it more suitable for practical applica-tions. For PLD growth, it is known that too low growth temperature will degrade the crystallinity. Further, previous experiment indicated that (0 0 1) orientated PZT is favored at low growth temperature which then will introduce impurity phase[20]. Therefore, 540 °C− 570 °C appears to be an optimized growth temperature. In addition, the presence of impurity phases affects the surface morphology as characterized by AFM. As shown inFig. 1c, the pure (1 1 1) phase has a dense grain and smooth surface with a roughness Rms of 3.0 nm. The film grown at 630 °C with a mixture of (1 0 1)/(1 1 0) and (1 1 1) phases has columnar structures with a gap between the columns. The roughness of thisfilm is 29 nm. The pure (1 0 1)/(1 1 0) phase grown at 680 °C shows a roughness of 6 nm.

Using these optimized growth conditions, PZT films with other compositions were grown as well.Fig. 2shows the XRDθ-2θ scans of 750 nm PZTfilms with different compositions grown on 0.5 nm MgO buffered GaN at 570 °C. As mentioned in our previous report, the MgO on GaN is stress free and a 0.5 nm MgO buffer layer is already thick enough for the epitaxial growth of PZT[13]. It was found that there are no impurity peaks for × ≥ 0.48 films, but the × < 0.48 films exhibit the mixture of (1 1 1) PZT and pyrochlore phase. The (1 1 1) PZT phase even fully disappeared for PZT (x = 0.2). Apparently, the Zr-rich favors the formation of impurity pyrochlore phase. It is consistent with a previous report that the perovskite transformation temperature is lower in Ti-rich PZT than in Zr-rich PZTfilms.[21]

In order to epitaxially grow Zr-rich PZT (x < 0.48), a 5 nm ul-trathin Ti-rich PZT (x = 0.8) was used as a seed layer for the growth of Zr-rich PZT. With these PZT (x = 0.8) seed layers, we were able to epitaxially grow all compositions of PZT on MgO buffered GaN as shown inFig. 3a. The XRD scan shows a pure PZT (1 1 1) phase for PZT (x = 0.2)films. More generally, all PZT compositions including Ti-rich PZT can be grown epitaxially on MgO/GaN by introducing a PZT (x = 0.8) seed layer (seeFig. 3a), e.g., PZT (x = 0.48). To further prove the epitaxial relationship, Phi-scans of PZT (4 2 0) and GaN (−1 2 4) peaks were performed (seeFig. 3b). The PZT peaks were found to be fully aligned with GaN (−1 2 4) peaks, demonstrating the epitaxial relationship of PZT on GaN. The surface morphology of the PZTfilms is shown inFig. 3c. Typical columnar structures reflected by the grain structure at the surface are densely arranged.

The ferroelectric properties of these 750 nm thick, epitaxial PZT thin films were investigated by measuring P–E (polarization versus electricfield) loops. Since a bottom electrode was absent, the P-E loop was measured using an in-plane configuration as described in our previous report[13]. The P-E loops were measured with a ferroelectric tester (AixAcct TF-2000 Analyzer) at a frequency of 1 kHz. Character-istic hysteretic P-E loops were observed for all PZTfilms with different compositions (seeFig. 4), demonstrating the ferroelectric properties of the PZTfilms. The remnant polarization (Pr) extracted from the P-E loop

for different compositions is shown inFig. 4b. The remnant polarization of PZT (x = 0.48) is about 17 µC/cm2_{and is found to be larger than}

PZT (x = 0.2) and PZT (x = 0.8). The coercivefield is found to be composition dependent as well (seeFig. 4c). The PZT (x = 0.8) pos-sesses the highest coercivefield of 108 kV/cm, about 4 times that of PZT (x = 0.48) and 2.7 times that of PZT (x = 0.2). The composition dependent coercivefield can be explained by the introduction of dif-ferent density of defect dipoles [22]. More defect dipoles are created with higher Zr/Ti disorder, e.g., at Zr/Ti ≈ 1, making the domain switching easier and therefore inducing lower coercivefield[22]. Due to the large coercivity, the PZT (x = 0.8) requires a much higher vol-tage to saturate the polarization (seeFig. 4a).

4. Conclusions

In summary, the growth of PZT on MgO/GaN is found to be com-position dependent. Pure epitaxial PZT (1 1 1) phase can be achieved for Ti-rich PZT which is located to the right side of the MPB. In contrast, the Zr-rich PZT which is located at the left side of the MPB is found to more easily form pyrochlore impurity phase. This composition depen-dent phase formation should be linked to the fact that the Ti-rich PZT has a lower perovskite transformation temperature than Zr-rich PZT [21]. The perovskite phase is more stable in Ti-rich PZT compounds. Regarding that the lattice constant of PZT which ranges from 4.0 Å to 4.1 Å depending on the composition × has relative large mismatch with MgO (2.6–5%), the introduction of PZT (x = 0.8) seed layer can relax the large strain from MgO. Additionally, the PZT seed layer can provides a perovskite surface template to guide the formation of per-ovskite phase. Therefore, the introduction of a Ti-rich perper-ovskite PZT as an additional buffer layer can greatly suppress the pyrochlore phase, leading to epitaxial growth of Zr-rich PZT.

Finally, the physical properties of these (1 1 1) PZT layers on GaN strongly depend on the composition. The PZT (x = 0.48) located at the MBP boundary has relative larger remnant polarization and lower coercivefield than the other compositions. The PZT (x = 0.8) is found to have the highest coercivefield. The different properties of PZT with different compositions can be used for different specific applications, such as high power FET’s and non-volatile memory ferroelectric con-trolled electronics device. Our strategy to realize epitaxial growth of PZTfilms with different compositions on GaN opens a new avenue for oxide-III-IV semiconductor electronics.

Declaration of Competing Interest

The authors declare that they have no known competingfinancial interests or personal relationships that could have appeared to in flu-ence the work reported in this paper.

Acknowledgements

Lin Li acknowledgesfinancial support from Nano Next NL (Grant no. 7B 04). The authors acknowledge NXP for providing the GaN/ AlGaN/Si (111) wafer.

References

[1] Tyler J. Flack, Bejoy N. Pushpakaran, Stephen B. Bayne, GaN Technology for Power Electronic Applications: A Review, J. Electron. Mater. 45 (2016) 2673. [2] S.J. Pearton, F. Ren, A.P. Zhang, G. Dang, X.A. Cao, K.P. Lee, H. Cho, B.P. Gila,

J.W. Johnson, C. Monier, C.R. Abernathy, J. Han, A.G. Baca, J.-I. Chyi, C.-M. Lee, T.-E. Nee, C.-C. Chuo, S.N.G. Chu, GaN electronics for high power, high temperature applications, Mater. Sci. Eng. B 82 (2001) 227.

[3] B. Jaffe, H. Jaffe, W.R. Cook, Piezoelectric ceramics, London: Academic Press. [4] Y.S. Kang, Q. Fan, B. Xiao, Y.I. Alivov, J.Q. Xie, N. Onojima, S.-J. Cho, Y.-T. Moon,

H. Lee, D. Johnstone, H. Morkoc, Y.-S. Park, Fabrication and current-voltage characterization of a ferroelectric lead zirconate titanate/field effect transistor, Appl. Phys. Lett. 88 (2006) 123508.

[5] S.K. Dey, S. Bhaskar, M.H. Tsai, W. Cao, Pb(Zr, Ti)O3-GaN Heterostructures for RF

MEMS Applications, Integr. Ferroelectr. 62 (2004) 69.

[6] W.P. Li, R. Zhang, Y.G. Zhou, J. Yin, H.M. Bu, Z.Y. Luo, B. Shen, Y. Shi, R.L. Jiang, S.L. Gu, Z.G. Liu, Y.D. Zheng, Z.C. Huang, Studies of metal–ferroelectric–GaN structures, Appl. Phys. Lett. 75 (1999) 2416.

[7] I. Stolichnov, L. Malin, P. Muralt, N. Setter, Nonvolatile gate effect in the PZT/ AlGaN/GaN heterostructure, J. Eur. Ceram. Soc. 27 (2007) 13.

[8] K. Elibol, M.D. Nguyen, R.J.E. Hueting, D.J. Gravesteijn, G. Koster, G. Rijnders, Integration of epitaxial Pb(Zr0.52Ti0.48)O3films on GaN/AlGaN/GaN/Si(111)

sub-strates using rutile TiO2buffer layers, Thin Solid Films 591 (2015) 66.

[9] B. Xiao, X. Gu, N. Izyumskaya, V. Avrutin, J.Q. Xie, H.Y. Liu, H. Morkoc, Structural and electrical properties of Pb(Zr, Ti)O3 grown on (0001) GaN using a double

PbTiO3/PbO bridge layer, Appl. Phys. Lett. 91 (2007) 182908.

[10] Y.R. Li, J. Zhu, W.B. Luo, Study of the integrated growth of dielectricfilms on GaN semiconductor substrates, IEEE T Ultrason Ferr. 57 (2010) 2192.

[11] E.A. Paisley, H.S. Craft, M.D. Losego, H. Lu, A. Gruverman, R. Collazo, Z. Sitar, J.-P. Maria, Epitaxial PbxZr1-xTiO3on GaN, J. Appl. Phys. 113 (2013) 074107.

GaN/sapphire by metalorganic chemical vapor deposition, J. Mater. Res. 21 (2006) 1562.

[13] L. Li, Z.L. Liao, N. Gauquelin, M.D. Nguyen, R.J.E. Hueting, D.J. Gravesteijn, I. Lobato, E.P. Houwman, S. Lazar, J. Verbeeck, G. Koster, G. Rijnders, Epitaxial Stress-Free Growth of High Crystallinity Ferroelectric PbZr0.52Ti0.48O3on GaN/

AlGaN/Si (111) Substrate, Adv. Mater. Inter. 5 (2017) 1700921.

[14] H. Hanawa, K. Horio, Increase in breakdown voltage of AlGaN/GaN HEMTs with a high-k dielectric layer, Phys. Status Solidi A 211 (2014) 784.

[15] N. Zhang, H. Yokota, A.M. Glazer, Z. Ren, D.A. Keen, D.S. Keeble, P.A. Thomas, Z.-G. Ye, The missing boundary in the phase diagram of PbZr1−xTixO3, Nature

Commun. 5 (2014) 5231.

[16] Y.L. Zhu, S.J. Zheng, X.L. Ma, L. Feigl, M. Alexe, D. Hesse, I. Vrejoiu,

Microstructural evolution of [PbZrxTi1– xO3/PbZryTi1– yO3] n epitaxial multilayers

(x/y = 0.2/0.4, 0.4/0.6)–dependence on layer thickness, Philos. Mag. 90 (2010) 1359.

[17] J.C. Agar, A.R. Damodaran, M.B. Okatan, J. Kacher, C. Gammer, R.K. Vasudevan, S. Pandya, L.R. Dedon, R.V.K. Mangalam, G.A. Velarde, S. Jesse, N. Balke,

A.M. Minor, S.V. Kalinin, L.W. Martin, Highly mobile ferroelastic domain walls in compositionally graded ferroelectric thinfilms, Nat. Mater. 15 (2016) 549. [18] A.R. Zomorrodian, A. Messarwi, N.J. Wu, AES and XPS study of PZT thinfilm

de-position by the laser ablation technique, Ceram. Int. 25 (1999) 137.

[19] J.-G. Ramí rez, A. Cortes, W. Lopera, M.E. Gó mez, P. Prieto, Scaling Laws in PZT Thin Films Grown on Si (001) and Nb-Doped SrTiO3(001) Substrates, Braz. J. Phys.

36 (2006) 1066.

[20] H.W. Song, The Effect of Deposition Temperature of Pb(Zr, Ti)O3 (PZT) Thin Films

with Thicknesses of around 100 nm on the Piezoelectric Response for Nano Storage Applications, World J. Condens. Matter Phys. 2 (2012) 51.

[21] C.L. Sun, J.J. Hsu, S.Y. Chen, A. Chin, Effect of Zr/Ti Ratios on Characterization of Pb(ZrxTi1-x)O3Thin Films on Al2O3Buffered Si for One-Transistor Memory

Applications, J. The Electrochem. Soc. 150 (3) (2003) G187.

[22] J.Y. Jo, S.M. Yang, H.S. Han, D.J. Kim, W.S. Choi, T.W. Noh, Composition-depen-dent polarization switching behaviors of (111)-preferred polycrystalline Pb (ZrxTi1−x)O3thinfilms, Appl. Phys. Lett. 92 (2008) 012917.